AILobstersabout 2 hours ago
Researchers Launch AIR-LLM for Edge AI
17 min read
Researchers present AIR-LLM, an architecture broadcasting LLM weights over radio for edge inference. Testing on LLaMA-3.1-8B shows 4.0% perplexity degradation. The system cuts energy by 157.7 times versus FP16 baselines and reduces airtime by 104.1 times with 20 users.
Level
Hype check
Tap to vote and see what everyone thinks.
